Intelligent identification and conversion method, device and equipment of mechanical design drawing and medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA CONSTR THIRD ENG BUREAU INSTALLATION ENG CO LTD
- Filing Date
- 2026-03-10
- Publication Date
- 2026-05-29
Smart Images
Figure CN122116406A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image recognition, specifically relating to a method, apparatus, equipment, and medium for intelligent recognition and conversion of mechanical design drawings. Background Technology
[0002] In the field of mechanical design and manufacturing, mechanical design drawings are crucial tools for conveying detailed design information. Traditionally, these drawings were hand-drawn, but with the development of computer-aided design (CAD) technology, the creation and editing of drawings have become more digital and automated. However, efficiently converting a large number of old-style or paper-based drawings into digital formats for further editing and analysis remains a challenge. Furthermore, even with digital drawings, automatically extracting key information and converting it into usable data formats, such as CAD models, is a significant technical challenge.
[0003] Currently, existing solutions mainly focus on image processing and pattern recognition technologies. These technologies include preprocessing drawing images using various algorithms, such as image sharpening, noise removal, and edge enhancement, to improve the readability of the drawings and the accuracy of subsequent processing. In terms of feature recognition, existing methods largely rely on traditional pattern recognition techniques, such as edge detection algorithms and feature matching, to identify specific elements in the drawings.
[0004] However, these traditional methods often have limitations when dealing with complex or low-quality drawings. For example, they may struggle to accurately identify blurry or obscured drawing elements, or they may not be able to effectively handle multiple different types of drawings. Furthermore, most existing methods require significant manual intervention when converting identified features into CAD models, limiting the degree of automation in the entire process.
[0005] It is evident that existing technologies have several major drawbacks in the automated recognition and conversion of mechanical design drawings: (1) Limited recognition accuracy: Traditional image processing and pattern recognition techniques are limited in accuracy and robustness when dealing with complex, blurry, or damaged drawings. These methods are difficult to effectively handle subtle differences and complex features on drawings.
[0006] (2) Lack of automation and flexibility: Many existing methods require a lot of manual intervention when converting identified drawing features into CAD models, which limits the degree of automation in the process. This not only reduces work efficiency but also increases the possibility of errors.
[0007] (3) Poor adaptability: Existing technologies usually work well for specific types or formats of drawings, but their performance and adaptability are poor for diverse and non-standardized drawings, such as old drawings or hand-drawn drawings. Summary of the Invention
[0008] The purpose of this invention is to address the problems existing in the prior art by providing a method, device, equipment, and medium for intelligent recognition and conversion of mechanical design drawings. By using a deep learning method based on YOLOv5, combined with advanced image preprocessing technology and automated CAD modeling technology, more accurate, automated, and flexible recognition and conversion of mechanical design drawings can be achieved.
[0009] According to one aspect of the present invention, a method for intelligent recognition and conversion of mechanical design drawings is provided, comprising: Perform image preprocessing on the mechanical design drawings to be processed; The preprocessed mechanical design drawings are input into the trained deep learning model for image recognition and feature extraction. The extracted features are transformed using an online modeling and transformation process to obtain a transformed computer-aided design (CAD) model.
[0010] The above technical solution, through YOLOv5-based deep learning, combined with advanced image preprocessing technology and automated CAD modeling technology, not only improves the accuracy of feature recognition, especially when processing complex or irregular drawings, but also automatically converts the recognized features into CAD models, greatly reducing the need for manual intervention and improving efficiency and accuracy.
[0011] As a further technical solution, after obtaining the converted computer-aided design CAD model, it also includes: optimizing the model based on computer-aided design CAD software and user feedback, and realizing a model correction and continuous learning mechanism.
[0012] As a further technical solution, image preprocessing is performed on the mechanical design drawings to be processed, including: sequentially performing Gaussian blurring, sharpening, and standardization on the mechanical design drawings to be processed.
[0013] As a further technical solution, the training of the deep learning model includes: Obtain mechanical design drawing samples of different types and annotate them with features of interest to create training and testing datasets; Input the drawing samples from the training dataset into the YOLOv5 model, perform edge detection and feature vector creation in sequence, and use the created feature vectors to predict the target, outputting the bounding box and class probability of the predicted target; The prediction results are validated using a test dataset until the test accuracy reaches a preset value, thus obtaining the trained deep learning model.
[0014] As a further technical solution, the extracted features are transformed using an online modeling and transformation process, including: forming a data matrix from all the extracted features, and performing two-dimensional or three-dimensional computer-aided design (CAD) modeling on the data matrix.
[0015] As a further technical solution, the method further includes: parsing the feature data output by the deep learning model; performing triangulation based on the parsed data to generate basic geometric shapes and structures; and generating a computer-aided design CAD model in a computer-aided design CAD system based on the generated basic geometric shapes and structures.
[0016] According to one aspect of the present invention, an intelligent recognition and conversion device for mechanical design drawings is provided, comprising: The preprocessing module is used to perform image preprocessing on the mechanical design drawings to be processed; The image recognition and feature extraction module is used to input the preprocessed mechanical design drawings into the trained deep learning model for image recognition and feature extraction. The conversion module is used to convert the extracted features using an online modeling conversion process to obtain the converted computer-aided design (CAD) model.
[0017] As a further technical solution, the device also includes: a model correction and continuous learning module, used to optimize the model based on computer-aided design CAD software and user feedback, and to realize a model correction and continuous learning mechanism.
[0018] According to one aspect of the present invention, an electronic device is provided, comprising: at least one processor, at least one memory, and a communication interface; wherein the processor, memory, and communication interface communicate with each other; the memory stores program instructions executable by the processor, and the processor invokes the program instructions to perform the method described thereon.
[0019] According to one aspect of the present invention, a non-transitory computer read storage medium is provided, the non-transitory computer read storage medium storing computer instructions that cause the computer to perform the method described herein.
[0020] Compared with the prior art, the beneficial effects of the present invention are: Improved recognition accuracy: By applying Yolov5-based deep learning technology, the recognition accuracy of various complex features on mechanical design drawings is significantly improved, especially when dealing with low-quality or complex drawings.
[0021] A highly automated conversion process has been achieved: by integrating advanced image processing technology and automated CAD modeling, the need for manual intervention is reduced, thereby improving the efficiency and accuracy of the entire conversion process.
[0022] The system's adaptability and flexibility have been enhanced: enabling the system to effectively handle various types and formats of drawings, including traditional hand-drawn drawings and old drawings, to meet a wider range of application needs.
[0023] Based on the aforementioned advantages, this invention overcomes the limitations of existing technologies and provides a more advanced and practical drawing processing solution for the field of mechanical design. Attached Figure Description
[0024] Figure 1 This is a flowchart illustrating the intelligent recognition and conversion method for mechanical design drawings according to an embodiment of the present invention.
[0025] Figure 2 This is a schematic diagram of the drawing preprocessing process according to an embodiment of the present invention.
[0026] Figure 3 This is a schematic diagram of the deep learning model workflow according to an embodiment of the present invention.
[0027] Figure 4 This is a schematic diagram of the online modeling and conversion process according to an embodiment of the present invention.
[0028] Figure 5 This is a schematic diagram of the model calibration and continuous learning process in an embodiment of the present invention.
[0029] Figure 6 This is a schematic diagram of the intelligent recognition and conversion device for mechanical design drawings according to an embodiment of the present invention.
[0030] Figure 7 This is a schematic diagram of the electronic device structure according to an embodiment of the present invention. Detailed Implementation
[0031] It should be noted that: This invention aims to solve key technical problems in the automated recognition and conversion of mechanical design drawings. Specifically, it includes: how to effectively perform high-quality image preprocessing on mechanical design drawings to ensure image clarity and standardization for subsequent processing; how to accurately utilize advanced deep learning technology (especially the Yolov5 algorithm) to identify and extract key features on the drawings, including parts, dimension lines, and text; how to efficiently convert the identified two-dimensional features into two-dimensional or three-dimensional CAD models, including estimation of element spatial positions and dimensions; and how to perform correction and continuous learning after model generation to ensure the model's accuracy and adaptability, especially when dealing with diverse and complex mechanical design drawings. Solving these technical problems will greatly improve the efficiency and accuracy of mechanical design drawing processing, providing strong support for engineering design and production.
[0032] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0033] This invention proposes an intelligent recognition and conversion method for mechanical design drawings, which mainly includes: (1) performing efficient image sharpening and standardization processing on the mechanical design drawings, and optimizing the image quality by combining Gaussian blur and sharpening algorithms to prepare for subsequent processing. (2) applying a deep learning model based on Yolov5 to perform drawing recognition and feature extraction, accurately identifying key information such as parts, dimension lines, and text on the drawings. (3) through an online modeling and conversion process, efficiently converting the identified features into two-dimensional or three-dimensional CAD models, including accurate estimation of the spatial position and size of elements.
[0034] In addition, it also includes: (4) implementing a model correction and continuous learning mechanism to optimize the model based on CAD software and user feedback to ensure its adaptability and accuracy.
[0035] like Figure 1 As shown, the intelligent recognition and conversion method for mechanical design drawings of the present invention includes: Step 1: Perform image preprocessing on the mechanical design drawings to be processed.
[0036] Image preprocessing, also known as image sharpening and standardization, is achieved by combining Gaussian blur and sharpening algorithms. This step first uses Gaussian blur to smooth the mechanical design drawing image, reducing noise and unnecessary details, making the image more suitable for subsequent processing. Then, a sharpening algorithm is applied to enhance the image's edges and details, improving image clarity and making important features on the drawing more prominent and easily identifiable. This combined approach balances image clarity and noise levels, providing optimized input for subsequent drawing recognition and feature extraction.
[0037] Furthermore, in the sharpening and standardization process, sharpening not only includes sharpening and noise reduction, but may also involve image contrast adjustment and color correction to ensure that all elements on the drawing are clearly visible. Standardization involves converting the image to a specific size and format, such as a uniform resolution and image format, to ensure that the image has consistent quality and format when input into the YOLOv5 model.
[0038] Step 2 involves inputting the preprocessed mechanical design drawings into the trained deep learning model for image recognition and feature extraction. This specifically includes: Data annotation: Before training the Yolov5 model, a large number of mechanical design drawings need to be accurately annotated, including marking key features such as parts, dimension lines, and text on the drawings. This process is crucial to ensuring that the model can learn how to correctly identify these features.
[0039] Model training and validation: The Yolov5 model is trained using labeled data to enable it to identify and locate key features on drawings. The model also needs to be validated during training to ensure it has good generalization ability and high accuracy.
[0040] Feature recognition and extraction: The trained Yolov5 model is used to analyze new mechanical design drawings, accurately identify key features on the drawings, and extract relevant information such as location and dimensions.
[0041] Step 3: Use the online modeling conversion process to convert the extracted features to obtain the converted computer-aided design CAD model.
[0042] Optionally, 2D CAD modeling involves converting the feature information identified by the Yolov5 model into 2D CAD primitives to create a 2D CAD model. This step involves converting elements such as lines, shapes, and text on the drawing into a format that CAD software can recognize and process.
[0043] Optionally, 3D CAD modeling involves further analyzing and processing the identified features of drawings that need to be converted into 3D models, and estimating their position and dimensions in 3D space. This includes resolving different views on the drawings and using geometric reconstruction techniques to create the 3D model.
[0044] The above method integrates image processing, deep learning, and CAD modeling technologies to improve the automation level and efficiency of mechanical design drawing processing, while ensuring the high quality and reliability of the processing results.
[0045] In embodiments based on the above method, such as Figure 2 As shown, the algorithm-based image sharpening and normalization process includes the following steps: Step 1.1: Collect photos of the mechanical design drawings using a camera and then transmit and store the data using a computer.
[0046] Step 1.2, Gaussian Blur, specifically involves calculating the Gaussian blur result for each pixel in the image by taking a weighted average of its surrounding pixels. This weighted average is based on a Gaussian function, which has the following form:
[0047] in, Is the Gaussian function at point... The value, It is the standard deviation, which controls the degree of ambiguity. The larger the value, the greater the degree of ambiguity.
[0048] When applying Gaussian blur, a window (usually a square) centered on the current pixel is used, and the average value of each pixel within that window is calculated using a Gaussian function. The formula can be expressed as:
[0049] in, It is the position after blurring. pixel values on The original image is in position pixel values, It is half the size of the window.
[0050] Step 1.3, Sharpening Process: Apply image sharpening algorithms to enhance edges and details, expressed by the following formula:
[0051] in, The output image of the sharpening algorithm, This indicates the original image. The image after Gaussian blurring. It refers to the sharpening intensity.
[0052] Step 1.4, format standardization, converts the images to a uniform format and size. For example, unify them to a consistent size and PNG format to improve the recognition accuracy of the subsequent Yolov5 algorithm.
[0053] In embodiments based on the above method, drawing recognition and feature extraction based on Yolov5 includes the following steps: Step 2.1: Based on the different types of mechanical design drawing sample data in Step 1, mark the features of interest (such as parts, dimension lines, text, etc.) on the drawing samples to create a YOLO model training and testing dataset. Step 2.2: Train the YOLOv5 model using the labeled drawing samples and test its performance on the validation set. Then, use the trained YOLOv5 model on new drawing images. Input the drawing image into the model, and the model will output the location and category of the recognized features.
[0054] When processing technical drawing recognition and feature extraction, the Yolov5 model first takes the digital image of the drawing as input and extracts features from basic to complex layers through a multi-layer convolutional network. These convolutional layers are equipped with different filters and activation functions such as ReLU are used to increase non-linear processing capabilities, thereby recognizing various details in the image, such as lines, text, and symbols. Subsequently, batch normalization layers optimize the training process and accelerate convergence, while feature fusion techniques combine features from different levels to improve the model's accuracy in recognizing drawing details and contextual information. Finally, the model's output layer transforms this information into specific prediction results, including bounding boxes, categories, and confidence scores for components, thus achieving accurate drawing element recognition and classification, supporting subsequent CAD modeling or other technical applications. This process not only improves recognition accuracy but also greatly enhances the model's adaptability and flexibility when processing complex drawings.
[0055] It should be noted that in the YOLOv5 model, for object detection, YOLOv5 uses a series of convolutional layers, residual connections, and upsampling layers to extract features and finally outputs the bounding box and class probability of each object.
[0056] The main mathematical formulas involve the forward propagation of neural networks, a key part of which is the convolution operation used for feature extraction:
[0057] in, It is the position after the convolution operation. The pixel values of the output feature map. These are the weights of the convolution kernel at position (m,n). denoted as the pixel value at position (i+m, j+n) of the input feature map, where a and b are half the size of the convolution kernel.
[0058] In embodiments based on the above method, such as Figure 3 As shown, the drawing recognition and feature extraction based on Yolov5 further includes: Edge detection is used to detect lines and shapes in an image.
[0059] The formulas involved in edge detection are:
[0060] in It is edge strength. and It is the gradient of the image in the horizontal and vertical directions.
[0061] In the YOLOv5 model, gradients are primarily used to identify edges in the horizontal and vertical directions of an image, which can be achieved through edge detection techniques.
[0062] Specifically, these operators detect edges by calculating the gradient intensity of image pixels:
[0063] in, f It is an image. and These are the gradients in the horizontal and vertical directions, respectively. The total magnitude of the gradient can be calculated from these two gradients. G and direction.
[0064] In embodiments based on the above method, the Yolov5-based drawing recognition and feature extraction further includes: feature vector creation. In each region of the image, based on gradient information (the result of edge detection), the neural network extracts key features and encapsulates them into feature vectors. These vectors describe key attributes of objects in the image, such as position, size (width and height), and angle.
[0065] Specifically, the creation of feature vectors transforms the detected features into a mathematical representation:
[0066] in It is an eigenvector. x,y These are the center coordinates of the target. w,h It refers to the width and height of the target. It refers to the direction or angle of the target.
[0067] In embodiments based on the above method, the drawing recognition and feature extraction based on Yolov5 further includes: generating bounding boxes and class probabilities.
[0068] Through further processing by the neural network, the aforementioned feature vectors are used to predict the bounding box and class probability of the target. In the Yolov5 model, especially in the last few layers of the network, feature vectors are used to predict the bounding box of each detected object (defined as...). This includes the probability of each category. This allows the model to not only identify objects in an image, but also determine their exact location, size, and probability of classification.
[0069] It should be noted that the entire process, from edge detection to the final bounding box and class probability output, is a series of consecutive steps. Specifically: 1) Edge detection provides preliminary image structure information, helping the model locate key regions in the image; 2) Feature vectors further refine the description of these regions, providing specific attributes for each region; 3) The bounding box and class probability are the final outputs, which are directly used in practical application scenarios, such as object detection and scene parsing.
[0070] In embodiments based on the above method, such as Figure 4 As shown, the online modeling transformation process includes the following steps: Step 3.1: Combine all extracted features into a data matrix and perform standardization processing.
[0071] Step 3.2: Convert the data matrix into a two-dimensional or three-dimensional CAD model.
[0072] 2D CAD modeling: Converting features into 2D CAD primitives, that is, the steps mentioned above. This is directly converted into parameters for CAD elements. The specific steps are as follows: (1) Data Preprocessing and Parsing: The initial data usually comes from the bounding boxes and class probabilities output by the Yolov5 model. First, this data needs to be parsed and converted into a structured format, such as converting the bounding box coordinates of the detected objects into actual engineering dimensions. Gaussian filtering is used to smooth the image to reduce noise and unnecessary details, improving the accuracy of subsequent processing. Laplacian filtering is applied to enhance the edges and details of the image so that lines and markings in the drawing can be more clearly identified.
[0073] (2) Geometric modeling: Delaunay triangulation is performed on the analyzed data to generate basic geometric shapes and structures. These algorithms can effectively construct continuous geometric shapes from scattered data. For cases requiring the generation of 3D models, the Radial Basis Functions (RBF) surface reconstruction algorithm is applied to derive 3D surfaces from 2D images.
[0074] Specifically, in 3D CAD modeling, the spatial positions and dimensions of different elements are estimated to construct a 3D model. That is, the spatial and dimensional information of 2D features is used to reconstruct the 3D geometry.
[0075] The formula can be expressed as: ,in It is a point in three-dimensional space. It is a point on a two-dimensional image. It is multi-dimensional spatial information.
[0076] Mechanical design drawings typically contain multiple views (such as front view, side view, top view, etc.), each providing information about the object in a specific direction. By analyzing these views, the relative position and dimensions of each element in three-dimensional space can be determined.
[0077] (3) Generation and optimization of CAD models: Import geometric data into the CAD system using the APIs of AutoCAD software. These APIs allow for automated data input and model building, improving efficiency. Further optimize and refine the model within the CAD software, such as adjusting alignment, dimensional accuracy, and adding necessary engineering attributes.
[0078] (4) Integration and Feedback Loop: Once the model is generated in the CAD software, it undergoes practical application testing, such as structural analysis or simulation, to verify the model's practicality and accuracy. Based on the test results and user feedback, the model parameters are adjusted or the algorithm settings are optimized.
[0079] After completing the above process, save the two-dimensional or three-dimensional CAD model on the computer.
[0080] In embodiments based on the above method, the intelligent recognition and conversion method for mechanical design drawings further includes: Step 4: Model calibration and continuous learning.
[0081] It's important to note that model calibration involves correcting the generated CAD model based on CAD software and user feedback. This may include adjusting dimensions, shape, or position to ensure the model's accuracy and usability. This step is crucial for guaranteeing the quality and suitability of the final model.
[0082] Continuous learning and model optimization involve constantly collecting user feedback and new drawing data to update and optimize the Yolov5 vision model, enhancing its accuracy and adaptability in drawing recognition. This ensures the system can adapt to constantly changing design standards and new types of drawings.
[0083] In embodiments based on the above method, such as Figure 5 As shown, the model calibration and continuous learning include: Step 4.1: Make corrections and adjustments based on CAD software and user feedback. or Parameters are set to match user feedback.
[0084] Model calibration based on CAD software and user feedback is an iterative process involving model evaluation, modification, and verification. First, feedback on the CAD model needs to be collected from actual users and design experts, including detailed comments on the model's accuracy, functionality, and existing problems. This feedback is then used in conjunction with performance evaluations performed using built-in analysis tools in the CAD software, such as finite element analysis (FEA), to identify and pinpoint specific problems in the model.
[0085] Secondly, based on the analysis results and user feedback, necessary adjustments are made to the CAD model. This may include changing design parameters such as dimensions, scale, or material properties, or making broader modifications to the design itself, such as rearranging components or reinforcing structural support. After these adjustments, the model needs to be updated in the CAD software, and the same testing procedures are used again to verify the effects of the modifications, ensuring that all adjustments effectively address the original problems.
[0086] The entire calibration process may require multiple iterations, each adjusted and verified based on the latest feedback. Simultaneously, all modifications and test results should be meticulously documented for future reference. Ultimately, this process ensures that the CAD model is not only technically accurate but also fully meets the user's actual needs and industry standards, thereby improving design reliability and product market adaptability.
[0087] Step 4.2: Continuously update Step 2, update the Yolov5 model, and improve recognition accuracy.
[0088] In summary, this invention significantly improves upon existing technologies in terms of recognition accuracy, automation, adaptability, and continuous optimization, and can more effectively meet the drawing processing needs of modern mechanical design. Specifically: Higher recognition accuracy: Utilizing a Yolov5-based deep learning model, this invention can more accurately identify complex features on mechanical design drawings, such as parts, dimension lines, and text. This method is particularly suitable for processing blurry or low-quality drawings, and compared to traditional pattern recognition techniques, it can more effectively handle subtle differences and complex features in drawings.
[0089] High degree of automation: The automated process of this invention significantly reduces the need for manual intervention, especially in the process of converting identified features into CAD models. This not only improves the efficiency of the entire process but also reduces the possibility of human error.
[0090] Continuous learning and optimization: Through a continuous learning mechanism, this invention can continuously optimize the recognition model based on user feedback and new drawing data, thereby improving accuracy and adaptability. This ensures that the technical solution continues to improve over time, adapting to new design standards and requirements.
[0091] Cost-effectiveness: The automated and efficient processing reduces labor costs while improving work efficiency. In long-term use, this invention can save users significant time and resources, especially in scenarios involving the processing of large numbers of drawings.
[0092] The implementation of the various embodiments of the present invention is based on programmed processing through a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of the present invention are encapsulated into various modules. Based on this reality, and building upon the above embodiments, the embodiments of the present invention provide an intelligent recognition and conversion device for mechanical design drawings, which is used to execute the intelligent recognition and conversion method for mechanical design drawings in the above method embodiments. See also... Figure 6 The device includes: a preprocessing module for preprocessing mechanical design drawings; an image recognition and feature extraction module for inputting the preprocessed mechanical design drawings into a trained deep learning model for image recognition and feature extraction; and a conversion module for converting the extracted features using an online modeling conversion process to obtain a converted computer-aided design (CAD) model.
[0093] The intelligent recognition and conversion device for mechanical design drawings provided in this embodiment of the invention adopts... Figure 6 Several modules within the system, based on deep learning models and combined with advanced image preprocessing and automated CAD modeling technologies, not only improve the accuracy of feature recognition, but also automatically convert the recognized features into CAD models, especially when dealing with complex or irregular drawings. This greatly reduces the need for manual intervention and improves efficiency and accuracy.
[0094] It should be noted that the device embodiments provided by the present invention, in addition to implementing the methods in the above method embodiments, are also used to implement the methods in other method embodiments provided by the present invention. The difference lies only in the setting of corresponding functional modules. Their principles are basically the same as those of the above device embodiments provided by the present invention. Anyone skilled in the art, based on the above device embodiments and referring to the specific technical solutions in other method embodiments, can obtain corresponding technical means and technical solutions constituted by combining technical features, and improve the device in the above device embodiments while ensuring the practicality of the technical solutions, to obtain corresponding device-type embodiments for implementing the methods in other method-type embodiments. For example: Based on the above-described device embodiments, as an optional embodiment, the intelligent recognition and conversion device for mechanical design drawings provided in this embodiment of the invention further includes: The model calibration and continuous learning module is used to optimize the model based on computer-aided design CAD software and user feedback, and to implement the model calibration and continuous learning mechanism.
[0095] Furthermore, the model calibration and continuous learning module is also used for: Corrections and adjustments were made based on CAD software and user feedback. or Parameters are adjusted to match user feedback; We continuously update the image recognition and feature extraction steps, update the Yolov5 model, and enhance recognition accuracy.
[0096] Based on the above-described device embodiments, as an optional embodiment, the intelligent recognition and conversion device for mechanical design drawings provided in this embodiment of the invention further includes: The first preprocessing submodule is used to smooth the image of the mechanical design drawing using a Gaussian blur algorithm to reduce noise and unnecessary details, making the image more suitable for subsequent processing. The second preprocessing submodule is used to apply sharpening algorithms to enhance the edges and details of the image, improve the image clarity, and make the important features on the drawing more prominent and easier to identify. The third preprocessing submodule is used to convert images to specific sizes and formats to ensure that images have consistent quality and format when input into the YOLOv5 model.
[0097] Based on the above-described device embodiments, as an optional embodiment, the intelligent recognition and conversion device for mechanical design drawings provided in this embodiment of the invention further includes: The first image recognition and feature extraction submodule is used for edge detection of lines and shapes in images; The second image recognition and feature extraction submodule is used to extract key features based on the edge detection results and encapsulate them into feature vectors. The third image recognition and feature extraction submodule is used to predict the bounding box and class probability of the target based on the feature vector.
[0098] Based on the above-described device embodiments, as an optional embodiment, the intelligent recognition and conversion device for mechanical design drawings provided in this embodiment of the invention further includes: The first transformation submodule is used to assemble all extracted features into a data matrix and perform standardization processing. The second conversion submodule is used to convert the data matrix into a two-dimensional or three-dimensional CAD model.
[0099] The method in this embodiment of the invention is implemented using an electronic device; therefore, it is necessary to introduce the relevant electronic device. For this purpose, embodiments of the present invention provide an electronic device, such as... Figure 7As shown, the electronic device includes: at least one processor, a communication interface, at least one memory, and a communication bus, wherein the at least one processor, the communication interface, and the at least one memory communicate with each other via the communication bus. The at least one processor invokes logical instructions stored in the at least one memory to execute all or part of the steps of the methods provided in the foregoing method embodiments.
[0100] Furthermore, when the logical instructions in at least one of the aforementioned memories are implemented as software functional units and sold or used as independent products, they are stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, is embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (a personal computer, server, or network device) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks—various media for storing program code.
[0101] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, located in one place, or distributed across multiple network units. The purpose of this embodiment is achieved by selecting some or all of the modules according to actual needs. Those skilled in the art will understand and implement this without any inventive effort.
[0102] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0103] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0104] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0105] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0106] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.
[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for intelligent recognition and conversion of mechanical design drawings, characterized in that, include: Perform image preprocessing on the mechanical design drawings to be processed; The preprocessed mechanical design drawings are input into the trained deep learning model for image recognition and feature extraction. The extracted features are transformed using an online modeling and transformation process to obtain a transformed computer-aided design (CAD) model.
2. The intelligent recognition and conversion method for mechanical design drawings according to claim 1, characterized in that, After obtaining the converted computer-aided design (CAD) model, the process also includes: optimizing the model based on the CAD software and user feedback, and implementing a model correction and continuous learning mechanism.
3. The intelligent recognition and conversion method for mechanical design drawings according to claim 1, characterized in that, Image preprocessing is performed on the mechanical design drawings to be processed, including sequentially performing Gaussian blur, sharpening, and standardization on the mechanical design drawings to be processed.
4. The intelligent recognition and conversion method for mechanical design drawings according to claim 1, characterized in that, The training of the deep learning model includes: Obtain mechanical design drawing samples of different types and annotate them with features of interest to create training and testing datasets; Input the drawing samples from the training dataset into the YOLOv5 model, perform edge detection and feature vector creation in sequence, and use the created feature vectors to predict the target, outputting the bounding box and class probability of the predicted target; The prediction results are validated using a test dataset until the test accuracy reaches a preset value, thus obtaining the trained deep learning model.
5. The intelligent recognition and conversion method for mechanical design drawings according to claim 1, characterized in that, The extracted features are transformed using an online modeling and transformation process, including: forming a data matrix from all extracted features, and performing two-dimensional or three-dimensional computer-aided design (CAD) modeling on the data matrix.
6. The intelligent recognition and conversion method for mechanical design drawings according to claim 5, characterized in that, The method further includes: parsing the feature data output by the deep learning model; performing triangulation based on the parsed data to generate basic geometric shapes and structures; and generating a computer-aided design CAD model in a computer-aided design CAD system based on the generated basic geometric shapes and structures.
7. An intelligent recognition and conversion device for mechanical design drawings, characterized in that, include: The preprocessing module is used to perform image preprocessing on the mechanical design drawings to be processed; The image recognition and feature extraction module is used to input the preprocessed mechanical design drawings into the trained deep learning model for image recognition and feature extraction. The conversion module is used to convert the extracted features using an online modeling conversion process to obtain the converted computer-aided design (CAD) model.
8. The intelligent recognition and conversion device for mechanical design drawings according to claim 7, characterized in that, The device further includes a model calibration and continuous learning module, used to optimize the model based on computer-aided design (CAD) software and user feedback, thereby implementing a model calibration and continuous learning mechanism.
9. An electronic device, characterized in that, include: The system includes at least one processor, at least one memory, and a communication interface; wherein the processor, memory, and communication interface communicate with each other. The memory stores program instructions that are executed by the processor, which invokes the program instructions to perform the method described in any one of claims 1 to 6.
10. A non-transitory computer read storage medium, characterized in that, The non-transitory computer read storage medium stores computer instructions, which cause the computer to perform the method described in any one of claims 1 to 6.